Download Dynamic PEM Fuel Cell Model for Power Electronics Member, IEEE,

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
1
Dynamic PEM Fuel Cell Model for Power Electronics
Design with Temperature Consideration
Ken Stanton, Member, IEEE, Jih-Sheng (Jason) Lai, Fellow, IEEE, and Douglas Nelson
Abstract—A dynamic model of a PEM fuel cell is developed for
power electronics simulation. The model accounts for static
losses of activation, ohmic, and concentration regions, and
dynamic transients due to charge double layer and compressor
delay. In this paper, study was given to additional factors,
including temperature, for their effects on the output voltage
with respect to load on the fuel cell stack. Findings were
analyzed and incorporated into the model based on testing with a
Ballard Nexa 1.2 kW PEM fuel cell unit. The model is developed
for PSPICE, then tested and compared to experimental data
from the Ballard fuel cell system.
Index Terms—Proton Exchange Membrane (PEM), fuel cell,
modeling, power electronics, dynamics.
F
I. INTRODUCTION
UEL cell systems are finding use in various power
applications, including automotive, residential, and
commercial. They show great promise, as they can utilize
hydrogen as fuel and produce water and heat as byproducts,
with relatively few moving parts and high efficiency
compared to combustion devices. The proton exchange
membrane fuel cell (PEMFC) has a low working temperature,
is compact and has good power density, and can respond
quickly to power demand changes, relative to other types of
fuel cells.
In most applications, power electronics are required in
order to use the PEMFC as the power source. The fuel cell is
considered a voltage source, with its output level depending
on power demanded of it. As such, applications that require a
constant or otherwise controlled voltage level must have
power conditioning electronics such as dc/dc converters.
Designing power electronics for fuel cell applications has
some challenges, including transient power handling and
system control with large variations in output voltage over the
operating range.
Whether or not such issues can be
effectively designed for depends highly on the quality of the
fuel cell model used in simulation. Software packages such as
Manuscript received xxx. This work is sponsored by the U.S. Department
of Energy (DOE) National Energy Technology Laboratory (NETL) SolidState Energy Alliance Program (SECA) under Award Number DE-FC2602NT41567. The Ballard Nexa 1.2kW PEMFC was provided by American
Power Converison Corp., whose donation allowed this research to happen.
Ken Stanton is a Ph.D. student with the Electrical and Computer
Engineering Department, Virginia Tech, Blacksburg, VA, 24060, USA (email: kstanton@vt.edu).
Jih-Sheng (Jason) Lai is a professor with the Electrical and Computer
Engineering Department, Virginia Tech, Blacksburg, VA, 24060, USA (email: laijs@vt.edu).
Fig. 1: Power electronics system driving ac load, powered by a fuel cell.
PSPICE, MATLAB, SIMULINK, and SABER accept the
power electronics and loads readily, so a fuel cell model is
needed that can be used with these programs as well. Fig. 1
shows an example fuel cell system to be simulated, including
the fuel cell, power electronics, and load.
A great number of fuel cell models exist, and for many
different types of simulation software. Extensive models can
be found in [1]-[6], all of which have fuel cell equations
directly entered into the model. Of these, [1], [2], and [4]
develop models in PSPICE, [2] and [6] in MATLAB/
SIMULINK, and [1] also in MathCAD; models in [3] and [5]
are not applied to any specific software packages. References
[7] and [8] develop fuel cell models using electrical
components and simple behavior blocks in PSPICE. Finally,
[9] and [10] explain fuel cell equations in great detail, discuss
modeling results, but don’t specifically show the models. All
papers above are for PEMFC’s except for [6], which is for a
solid-oxide fuel cell (SOFC).
The extensive models [1]-[6] provide very accurate output
voltage and currents, as well as providing outputs like
temperature and air humidity. However, these models are
slow to simulate, cumbersome to use, and require extensive
knowledge of inputs to the fuel cell. Electrical models in [7]
and [8] resolve these problems, but lose accuracy in the output
as the temperature and hydration of the fuel cell system are
not considered.
The work presented in this paper takes the model shown in
[8] and adds large time-constant dynamics, primarily due to
temperature. Temperature effects cannot be isolated, so other
factors including hydration, humidity, cooling system, and the
Fig. 2: Ballard NEXA 1.2kW PEM fuel cell system, including hydrogen
pressure regulator, air compressor, and control system with sensors. This
system was used for the development work in this paper.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
control system are inherently present. At a given load
condition, a rise in output voltage can be observed over a time
period of minutes, the net effect of which is addressed herein.
Additionally, the causes of these dynamics are studied in
detail, with a focus on the role of temperature. Development
testing was performed with a Ballard Nexa 1.2 kW PEMFC
system, shown in Fig. 2.
2
by properties of its operation. In the PEMFC, there is the
ideal voltage output, E, which is often called the open circuit
voltage 1 (OCV), or theoretical reversible voltage. This
voltage is calculated as
Ecell = E0,cell +
RT ⎡ *
ln p H × p*O ⎤ − Ed ,cell ,
2 ⎦
2F ⎣ 2
(3)
II. BASIC PEMFC OPERATION AND EQUATIONS
This section presents basic background information on fuel
cell operation, supported by equations. For this work, some
assumptions were made to simplify both equations and
analysis, as follows [2]:
1. All gases behave ideally and are distributed uniformly.
2. The fuel is humidified H2 and the oxidant is humidified
ambient air.
3. Thermodynamic properties are evaluated at the average
stack temperature, temperature variations across the stack
are neglected, and the overall specific heat capacity of the
stack is constant.
4. Parameters of individual cell performance can be lumped
together to represent a fuel cell stack.
A. Static Operational Conditions of PEMFC
For any given load condition in the operational range of a
PEMFC, certain factors govern the output; Fig. 3 shows this
graphically in the typical I-V fuel cell curve. These factors
are shown below, observed for a single fuel cell in a fuel cell
stack. The output voltage of a given fuel cell is
Vcell = Ecell − Vact ,cell − Vohm ,cell − Vconc ,cell ,
E0,cell = E 0,0cell − k E (T − 298) ,
(2)
where Ncell is the number of cells in the stack [2]. All other
nomenclature in (1) is defined below.
1) Open Circuit Voltage: A fuel cell can be considered a
voltage source, having an ideal voltage level which is reduced
(4)
where E00,cell is the reference potential at standard temperature
and pressure (298 K, 1 atm), and kE is a constant, both
positive.
⎛ t ⎞
Ed ,cell (t ) = λei (t )[1 − exp ⎜ − ⎟]
⎝ τe ⎠
(5)
is the final element of the OCV equation, where λe is a
constant, i(t) is the cell current, and τe is fuel and oxidant flow
delay, all positive.
2) Activation Polarization: Under low power demands, the
electrochemical reaction is slow at the electrode surface due to
the nature of its kinetics [7]. This loss of potential is
(1)
and therefore the stack output voltage is
Vout = Vcell × N cell = E − Vact − Vohm − Vconc ,
where R is a gas constant, T is temperature in Kelvin, F is
Faraday’s constant, and p* is the partial pressure of the noted
species, all of which are positive. Furthermore,
Vact = η0 + a(T − 298) + bT ln( I ) ,
(6)
where a, b, and η0 are constants, all positive [2]. Activation
polarization therefore has a natural log curve over the range of
current loads, as can be seen in Fig. 3.
3) Ohmic Polarization: This region is largely linear, as it
relates to loss of potential due to electrical resistance of the
polymer membrane, between the membrane and electrodes,
and in the electrodes themselves. The ohmic losses can be
expressed as
Vohm = IRohm
(7)
with
Rohm = Rohm 0 + k RI I − k RT T ,
(8)
which has respective k constants for both the current and
temperature dependent terms, and a constant term Rohm0, all
positive.
4) Concentration Polarization: This loss, also called massFig. 3. Typical fuel cell polarization, or I-V, curve. Ideal potential
(top) is reduced by three polarization regions (bottom three curves), to
the commonly seen result (second from top).
1
The terminology “open circuit voltage” is commonly used in fuel cell
literature, but can be confusing to those in electronics. Here it is meant as the
voltage level if no polarization losses were present.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
3
transportation loss, occurs at high current loads due to an
inability to supply reactants and remove products fast enough
for the chemical reaction to occur completely.
The
concentration polarization is defined as
Vconc = −
RT ⎛
I ⎞
ln ⎜1 −
⎟
zF ⎝ I lim ⎠
(9)
where z is the number of electrons participating, and Ilim is the
limitation current, both positive. The curve can be observed
in Fig. 3 by the severe drop-off at high current demand. It
should be noted that fuel cells are typically designed to avoid
operating in this region.
B. Dynamic Operating Conditions of PEMFC
Major short time-constant dynamic conditions can be
observed during load transients in Fig. 4. When a load step
occurs, the short-term fuel cell response is dominated by two
major phenomena: mass transport loss and compressor speed
delay. Mass transport delay occurs when there is a deficiency
(or excess) of reactants (products) at the reaction site. The
fuel cell voltage drops (rises) like a capacitor due to the
charge double layer effect. This effect is capacitor-like
because of the physical structure of the fuel cell electrodes
separated by the membrane. The voltage drop (spike) occurs
due to a delay in change of the compressor speed, which
supplies ambient air, and needs to adjust to the load demand.
In Fig. 4, the dip in fuel cell voltage shows both of these
phenomena, with the initial falling curve a result of mass
transport loss’s capacitance, and the recovery occurring as the
compressor comes to speed. These are both first order
responses, so for a time simulation, they can be modeled with
Vcdl = e -t /τ 2 -1
(10)
Vcomp = 1- e-t /τ1 ,
(11)
and
Fig. 5. Temperature (upper curve) and voltage (lower curve) data for a
full-load step transient, showing long time-constant rises in both. This
gave motivation to improve fuel cell models that do not incorporate
these long time transients.
where τ1 and τ2 are respective time constants for the charge
double layer discharge and compressor speed change. The
charge double layer time constant is on the order of a few
microseconds, and the compressor time constant around a
hundred milliseconds. The resulting dynamic transient is the
sum of (10) and (11).
The dynamic conditions are important for an accurate
model, as power electronics designers need to consider energy
storage if they wish to suppress these effects.
C. Additional Dynamics of PEMFC, Including
Temperature
In Fig. 5, a load step occurred, increasing from no-load to
full-load at time t=0. (Note: the dynamics discussed in section
B cannot be seen in this figure.) There is a voltage dynamic
present, on the order of many seconds, which was not
included in the previous sections. This transient is the focus
of the model improvements made in this paper.
At first, it appeared that this transient was solely influenced
by stack temperature. According to [2],
qnet = qchem − qelec − q sens + latent − qloss ,
(12)
where q refers to energy and q is power. Therefore net power
is the power released by the chemical reaction reduced by
electrical power, heat power due to reactants and products
flowing through the stack (sensible and latent), and power lost
to the surroundings (by heat). This net power is that
responsible for changes in fuel cell temperature, as given by
M FC CFC
Fig. 4. Short time-period fuel cell dynamic showing no- to full-load step.
Transient points of interest are circled on voltage and power curves.
dT
= qnet ,
dt
(13)
where MFC is the total mass of the fuel cell and CFC is its
specific heat capacity. Equations (12) and (13) could be used
to predict stack temperature if the terms of (12) could be
evaluated well and evaluated simply. However, the equations
behind those terms, as given in [2], require extensive
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
knowledge of fuel and air characteristics to be known, and
also depend on temperature themselves. As well, data from
tests with the Ballard fuel cell system support this notion, i.e.
a solid relationship between electrical characteristics and stack
temperature was simply not attainable. Therefore, if the
model being developed is to remain simple and reliable,
temperature prediction directly should not be attempted.
Despite this, it is still apparent that the voltage transient is
related to temperature. Fig. 4.2 in [1] shows increasing stack
temperature as generally causing an increase in output voltage,
i.e. it shifts the I-V curve upward. As well, [2], [3], [5], and
[9] state that the power consumed by static polarization losses
is considered the heat source for fuel cell temperature change,
and this power increases with stack output power demanded.
Therefore, in load step dynamics like that of Fig. 5, it can be
concluded that there is a direct relationship between changing
stack temperature, changing output power, and changing
output voltage.
Based largely on this, it was noted that as power demanded
of the fuel cell stack changed, so did the polarization losses.
These losses are the major contributors to stack heating and
cooling, so therefore power demanded of the stack is related
to stack temperature. As temperature changes, along with
other factors such as cell hydration, so does the magnitude of
the polarization losses, even at a constant load condition. As
these losses change, the output voltage of the stack is directly
affected. Therefore, it was concluded that there should be a
correlation between the power demanded of the fuel cell stack
and the stack’s output voltage, via changes related to
temperature. This hypothesis is the basis of the model
proposed herein, and was tested and developed. Section III
will display and discuss the results.
D. Other Operation Modes for Fuel Cell System
Over and above standard operation, the tested Ballard Nexa
fuel cell system has additional operating modes, including a
protective shutdown. These modes should be considered if
creating an extensive fuel cell system model.
1) Cold Start Operation: When started from a cold
condition (stack temperature less than or equal to ambient),
Ballard’s control system operates in cold start mode. This is a
protective mode that limits output power, approximately 300500 W, for 2 minutes. If the load attempts to exceed the
power level in the set time, the fuel system goes into
shutdown mode, suspending all activity.
2) Protective Shutdown: Protective shutdown of the fuel
cell system occurs whenever a measured parameter exceeds its
preset limit. These parameters include fuel pressure, stack
temperature, power output, and fuel (hydrogen) leakage. In
such an event, all output is stopped and fuel cell operation
ceased.
3) Rejuvenation Mode: Part of normal shutdown, the fuel
cell system disconnects itself from the output and runs
independently to control hydration of the stack. This mode
would not need to be modeled (unless including fuel
consumption), as it is part of normal shutdown and does not
produce any electrical output [11].
4
III. POWER BASED MODEL DEVELOPMENT
A. Data Collection and Analysis
Dozens of load step curves were captured and analyzed from
the Ballard Nexa fuel cell system via its serial PC connection.
The data were collected and analyzed in Fig. 6. Change in
voltage was calculated by subtracting the post-transient
voltage level from the results of a fast load sweep, which
would keep temperature relatively constant. A linear curve-fit
was then used to estimate the results, with only a few major
outliers. These outliers and other differences in values from
test to test can be attributed to:
1. Ambient variations including temperature, humidity,
air quality, and fuel cell containment
2. Fuel variations including purity and humidity
3. Hydration level of the stack
4. Age degradation of stack, or damage to it
5. Variations in sensors, compressor integrity, or other
physical variations of balance of plant components
6. Variations in the integrity of the load under test
The result presented here can be interpreted as such: if a
load step were performed from a steady-state no-load
condition, then the voltage can be expected to change by the
amount specified, after the large time-constant transient is
complete. For the Ballard Nexa system tested, the linear
curve-fit has a slope of 2.2 V/kW. Therefore, for a 1.2 kW
(full-load) step, the voltage will change 2.5 V from start to
finish of the transient. Since output voltage at this point is
28.5 V, this amounts to nearly a 10% change in voltage – a
very significant level to a power electronics designer.
B. Load Dependent Model
The fuel cell model in [8] includes all the static and dynamic
conditions discussed in section II A & B. The model
proposed here extended this work by adding the relationship
from the discussion in section III A.
Fig. 6. Stack output voltage change for a given load condition. This
general relationship is the basis for the power based model.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
Fig. 9. Fuel cell model with dynamic subsystem modification added.
Fig. 7. Fuel cell model with static modification subsystem added.
1) Static Modification: The static subsystem of the model
is established by calculating stack output power, as product of
current and voltage, and feeding this into a voltage-dependent
voltage-source (VDVS). The VDVS has a gain of 2.2 V/kW
as calculated, and is oriented such that positive stack power
creates an increase in the model’s output voltage. A
simplified block diagram of this system can be seen in Fig. 7.
Note that this diagram has no dynamic components.
When the static subsystem is implemented, a shift in the
static I-V curve occurs. Fig. 8 illustrates this. The lower,
darker curve was created by taking data directly from [11].
Following, the upper, lighter curve was created by adding the
linear 2.2 V/kW increase to the other curve, illustrating the
difference in steady-state output voltage.
2) Dynamic Modification: During a load-step transient,
output voltage changes exponentially over time. As stated
previously, this is primarily due to temperature changes, but
hydration, humidity, etc. can also affect this curve. The
dynamic being modeled has an exponential response over
time, so a LaPlace block is used to simulate it (block ES2).
The LaPlace block takes in the output of the static block and
applies it to
GLaPlace ( s ) =
1
,
τ s +1
5
(16)
where τ is the dynamic time constant, thereby creating an
exponential response to the transient. A block diagram of the
model with the dynamic subsystem can be seen in Fig. 9.
Fig. 8. I-V curve of Ballard Nexa fuel cell. Lower curve is original
output voltage curve, taken from [11]. Upper curve is output voltage with
the static subsystem in effect.
C. Setting Model Parameters for Ballard System
To use the proposed model for the Ballard Nexa fuel cell
system, only a single time constant needs to be determined.
The time constant must be chosen carefully, as it is highly
related to temperature, which can be altered by the Ballard
controls via the cooling fan and compressor. As a result,
small load step simulations will be highly accurate, as the
temperature controls are unlikely to change. Alternately, it
can be argued that large load steps include all of these effects,
and therefore are more representative of the system
holistically.
Testing on the Ballard system led to a fairly consistent
dynamic time constant of 50 seconds, for both large and small
load steps. This value is entered into (16) to complete the
model modification for the Ballard fuel cell system.
D. Fuel Cell Model Operation
The model demonstrated in this paper has many components
and component systems representing fuel cell phenomena.
These phenomena have been discussed in section II, including
static polarizations and dynamics such as charge double layer.
A description of how the fuel cell characteristics are modeled
with electrical components in Figs. 8 and 9 follows.
Additionally, values of these components are all shown in
Table 1.
Considering first the static operation of Fig. 8, the circuit has
two current flow paths. When the load current is low, the
Fig. 10. I-V curve from simulation of PSPICE model set up for Ballard
fuel cell system. Upper curve represents electrical model with proposed
modification in effect, and lower curve without. The lower curve closely
matches that found in [11] for the Ballard system.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
Fig. 11. Full-load step simulation using PSPICE model developed.
power flows through Q1, R3, and R4 to the load. With R4
voltage drop being low, transistor Q2 will be cut off. This
operating condition represents the activation polarization
region. As the load current increases, the R4 voltage
eventually exceeds the required base-emitter junction voltage,
and Q2 starts conducting. At this time, operation is in the
middle region of the I-V curve of Fig. 3, dominated by ohmic
losses. In this model, the combination of resistances R1, R3,
Fig. 12. Experimental test results for 4 load steps performed on Ballard
fuel cell system.
6
R4, and collector-emitter junction impedances model the
actual ohmic loss. Concentration polarization is not modeled
directly as the Ballard system (as well as most others) are
designed such that they never operate statically in this region.
The dynamic model of Fig. 9 employs two first-order
dynamic blocks, E2 and E3, to represent the load transient
effect. The first time constant (E3) is the charge double-layer
effect, which is in the order of sub-microsecond, and the
second time constant (E2) is the compressor delay, which is in
the order of sub-second. As a transient load applies, with a
nearly invisible 1st time constant delay, a load currentcontrolled voltage source (CCVS) quickly increases the
multiplier output to increase E1 drop, and thus the output
voltage drops. When the output voltage decreases, the
multiplier input also decreases with a delay related to the 2nd
time constant, the compressor delay. Consequently, the
multiplier output and the E1 drop decreases, thus bringing the
output back higher, which represents the effect of the air
compressor.
IV. MODEL VALIDATION
After completing the modified fuel cell model in PSPICE,
simulations were run to compare to experimental fuel cell test
results. The simulations were set up to match the timing and
magnitude of loads applied in the experimental tests.
1) I-V Curve: The first simulation is a load sweep from
zero to full-load current, with and without the model
modification active. The result is shown in Fig. 10, with the
lower curve representing the output without the proposed
dynamic modification, and upper with modification. This
result closely matches that of Fig. 8, the static modeling goal.
2) Dynamics: Fig. 11 shows the result of a full-load step
transient simulation, and can be compared to Fig. 5. At time
t=0, a load step occurred, increasing from no-load to full-load.
First, note that the voltage level of the simulation is higher
than the experimental result. This is largely attributed to age
of the Ballard fuel cell system, which has degraded in its
ability to output its full potential; other such factors were
listed in section III A.
Additionally, a simulation was run to compare to a series of
part-load steps performed on the Ballard system. The
experimental test started at no-load, stepped to 225 W, 360 W,
480 W, and 680 W for 450 s each, and then back to no-load;
the simulation followed the same pattern. Fig. 12 shows the
experimental result, and Fig. 13 shows the simulation. Again,
note that a slightly higher voltage level is observed in the
simulation, as mentioned previously. Also, a small drop in
voltage appears during the first load step in Fig. 12. It is
possible that this load condition was near the border of two
compressor or cooling fan speeds, and the speed changed
during this load condition test. This phenomenon was not
accounted for in the model developed, and hence will not
appear in Fig. 13.
V. CONCLUSION
Fig. 13. Simulation results from PSPICE model developed for he same
load conditions in Fig. 12.
Previous fuel cell models are either over-complicated or
over-simplified, making them poorly suited for many
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
applications. In this paper, an expanded electrical model of a
PEMFC was proposed and tested, and results compared to
experimental data. The results show that the proposed model
is more representative of actual fuel cell output than previous
simple electrical models discussed in the introduction. To
achieve this, the previous model was improved by adding a
temperature-related dynamic observed over a period of many
seconds. Also, the model is simpler to set up and use than the
more complicated models available previously. It can be used
as-is to simulate a Ballard Nexa 1.2 kW system, or its values
can be modified for another PEMFC using minimal test data.
Since the goal of this work was to improve electrical modeling
state-of-the-art while maintaining model simplicity and fast
simulation speed, the work was a success. The power
electronics designer can use this model to simulate an entire
system, such as that suggested in Fig. 1, and experience
improved representation of fuel cell output characteristics
over some previous models.
APPENDIX
TABLE I
MODEL COMPONENT VALUES
R1
R2
R3
Value /
Equation
0.036
0.01
0.32
R4
0.061
Ω
RP
RL
C1
E1
E2
25.0
0.65–inf
2.0
1
-0.75/
(1+1.2s)
0.05/
(1+0.12s)
0.06
-0.004
0.27
1/(1+50s)
0.16
0.16
100
10e6
45
Ω
Ω
uF
(gain)
(gain)
High current branch resistor
Model internal filter resistor
High current branch control resistor
1
High current branch control resistor
2
Parasitic Load Resistor
Load
Model internal filter capacitor
Transient voltage-drop block
Compressor delay modeling block
(gain)
Charge double layer modeling block
(gain)
(gain)
(gain)
(gain)
(gain)
(gain)
Area
Area
V
Output voltage scaling
Output current scaling
Subsystem VCVS
Subsystem dynamic LaPlace block
Fuel cell current CCVS
Load current CCVS
QbreakN setting
QbreakN setting
Ideal (open-circuit) fuel cell voltage
Symbol
E3
E4
E5
ES1
ES2
H1
H2
Q1
Q2
VOC
Units
Ω
Ω
Ω
Description
REFERENCES
[1]
[2]
[3]
Jerome Ho Chun Lee, “A Proton Exchange Membrane Fuel Cell
(PEMFC) Model for Use in Power Electronics,” M.S. thesis, Dept. of
Electronic and Information Eng., The Hong Kong Polytechnic Univ.,
Hung Hom, Hong Kong, 2004.
C. Wang, M. H. Nehrir, and S. R. Shaw, “Dynamic Models and Model
Validation for PEM Fuel Cells Using Electrical Circuits,” IEEE Trans.
Energy Conv., vol. 20, no. 2, pp. 442-451, June 2005.
P. Srinivasan, J. E. Sneckenberger, and A. Feliachi, “Dynamic Heat
Transfer Model Analysis of the Power Generation Characteristics for a
Proton Exchange Membrane Fuel Cell Stack,” in Proc. 35th Southeastern
Symp. on Sys. Theory, Mar. 2003, pp. 252-258.
[4]
7
P. Famouri and R. S. Gemmen, “Electrochemical Circuit Model of a
PEM Fuel Cell,” at Power Eng. Soc. Gen. Meeting, July 2003, pp. 14401444.
[5] X. Kong, A. M. Khambadkone, and S. K. Thum, “A Hybrid Model With
Combined Steady-state and Dynamic Characteristics of PEMFC Fuel
Cell Stack,” in Conf. Record 2005 Ind. App. Conf., Oct. 2005, pp. 16181625.
[6] K. Sedghisigarchi and A. Feliachi, “Dynamic and Transient Analysis of
Power Distribution Systems With Fuel Cells-Part I: Fuel-Cell Dynamic
Model,” IEEE Trans. Energy Conv., Vol. 19, no. 2, pp. 423-428, June
2004.
[7] D. Yu and S. Yuvarajan, “A Novel Circuit Model For PEM Fuel Cells,”
at Applied Power Elect. Conf. and Exp., 2004, pp. 362-366.
[8] Jih-Sheng Lai, “A Low-Cost DC/DC Converter for SOFC,” presented at
SECA Annual Review Meeting, Boston, MA, May 2004.
[9] W. Friede, S. Raёl, and B. Davat, “Mathematical Model and
Characterization of the Transient Behavior of a PEM Fuel Cell,” IEEE
Trans. Power Elects., Vol. 19, no. 2, pp. 1234-1241, Sep. 2004.
[10] J. M. Corrêa, F. A. Farret, L. N. Canha, and M. G. Simões, “An
Electrochemical-Based Fuel-Cell Model Suitable for Electrical
Engineering Automation Approach,” IEEE Trans. Ind. Elects., Vol. 51,
no. 5, Oct. 2004.
[11] Nexa Power Module User’s Manual, Ballard Power Systems Inc.,
Document Number MAN5100078, June 2003.